[ntp:questions] Re: Good GPS for attic?

David Schwartz davids at webmaster.com
Sat Feb 12 23:13:59 UTC 2005


"Richard B. Gilbert" <rgilbert88 at comcast.net> wrote in message 
news:BOOdnTvGHouOEpDfRVn-pA at comcast.com...

> It works as designed!  In the case you describe, you know what time it is 
> plus/minus 15 milliseconds.   If you configure four or five servers you 
> may get slightly better time.   If you need accuracy greater than +/- 
> 10-15 milliseconds, you are correct in that you probably can't get it from 
> the internet.   For many applications +/- 10-15 milliseconds is good 
> enough.

    And for many it's not. In fact, for a large class of applications (those 
involving benchmarking network operations), the error involved in the time 
setting process is exactly the time you were trying to measure in the first 
place!

> For those that need accuracy greater than this, GPS clocks are probably 
> the way to go.   Rubidium or Cesium clocks are also possibilities for 
> those with lots of money.

    GPS clocks with NTP can currently get you about three microsecond 
accuracy. That is, at any time, you know what time it is within three 
microseconds of UTC. Rubidium or cesium clocks actually won't increase this 
accuracy. However, they can provide greater stability (that is, you won't be 
+3uS one second and -3uS the next). Currently, the optimum price/performance 
point seems to be a Linux or FreeBSD machine running NTP with a GPS clock 
that provides a PPS output.

    DS





More information about the questions mailing list